2023, Vol. 10, No. 1. - go to content...

Permanent address of this page - https://resources.today/en/59ecor123.html

Метаданные этой статьи так же доступны на русском языке

DOI: 10.15862/59ECOR123 (https://doi.org/10.15862/59ECOR123)

Full article in PDF format (file size: 728.4 KB)

For citation:

Karavaev A.E. Intelligent tuberculosis recognition system using deep learning. Russian journal of resources, conservation and recycling. 2023; 10(1). Available at: https://resources.today/PDF/59ECOR123.pdf (in Russian). DOI: 10.15862/59ECOR123

Intelligent tuberculosis recognition system using deep learning

Karavaev Artyom Evgenievich
Financial University under the Government of the Russian Federation, Moscow, Russia
E-mail: karavaev_artem@mail.ru

Academic adviser: Chernyakov Aleksey Nikolaevich
Financial University under the Government of the Russian Federation, Moscow, Russia
E-mail: ANChernyakov@fa.ru

Abstract. In this scientific publication, the author discusses the problems associated with the diagnosis of tuberculosis and proposes a new approach to solving this problem. He proposes to use deep learning in intelligent TB recognition systems to improve the accuracy of diagnosis. The author analyzes in detail the advantages and limitations of a recognition system using deep learning, and also considers various technical and scientific problems associated with the creation of such a system. The author highlights the importance of using deep learning in intelligent TB recognition systems to increase diagnostic accuracy and improve treatment outcomes. He also discusses previous research in this area and draws conclusions about the promise of using deep learning to diagnose TB. The author proposes new approaches to improve the efficiency of the tuberculosis recognition system based on deep learning. He describes the training methods used to create such systems and discusses various aspects of their design, from data collection and processing to analysis of the results. One of the most promising, of course, is medicine. In this area, a doctor’s mistake can be fatal. Machine learning methods are able to minimize the mistakes of doctors. In this work, the main attention is paid to the tuberculosis recognition system. The aim of the work is to create an effective model of a neural network capable of recognizing tuberculosis with high accuracy, and also be useful in healthcare in conditions of limited resources. The main quality parameter of the created model is the recognition accuracy of the image class. Lung radiographs are images of a reduced dimension compared to the original materials, due to which high performance is achieved when they are processed by the convolutional network in the next stage. An extensive database is needed to achieve good results. However, by optimizing hyperparameters as well as an efficient network architecture, high accuracy rates can be achieved. The model consists of several convolutional layers, followed by fully connected layers, such an architecture allows you to get consistently high accuracy values on the test set.

Keywords: computer vision; artificial intelligence; tuberculosis recognition system; intelligent system; deep learning; percentage of correct recognitions; medicine

Download article in PDF format